More Evidence on Car Seats vs. Seat Belts

Things move quickly in the modern world. Within two hours of posting my academic paper on car seats vs. seat belts on the Freakonomics web page (the first time this paper had seen light of day), another economist found the paper and tested its hypotheses on a very different data set and reported back the results.

The economist is Paul Heaton, a graduate student at the University of Chicago and a former co-author of mine. (Our affiliation is largely irrelevant here — he just happened to have data handy to test my results — but I mention it in the interest of full disclosure). The data he has are from New Jersey motor vehicle crashes. The big difference between his data and mine are that he has all crashes (even if no one dies. I only had access to crashes in which someone died. This difference is important, because a concern in the fatal crash data is what economists call “sample selection.” The choices people make about what safety device to use will affect whether they die, which in turn affects which crashes I see in my data.

Heaton replicated the most basic specifications in my paper. His results are remarkably similar. He found no difference in the death rates or incapacitating injuries for children in car seats versus children using adult seat belts. Like me, he found a slight advantage for car seats in preventing non-incapacitating injuries relative to adult seat belts (car seats offered a 10% improvement for these less serious injuries in both of our samples). The only difference is that in his data the gap in injuries between car seats and seat belts is statistically significant, whereas in my data set it was not.

When I compare my findings to the existing NHTSA estimates on fatalities, I can see how our approaches differ and why we get different answers. What is more puzzling to me is why my results and Heaton’s both suggest very little injury benefit of car seats, but the medical literature often finds 70% (!!) reductions of injuries with car seats relative to seat belts. We find reductions that are an order of magnitude smaller. They use very different methods — surveying people in the weeks after crashes for instance — but still it is really a puzzle. Which is why, when you read my paper, I am extremely cautious in interpreting the injury findings.

I hope that the medical researchers, Heaton, and I can all work together to try to make some sense of the conflicting results being generated by these different methodologies to resolve this important question.

Chris Adams

Having almost died at the age of 4 (in 1973) when I was thrown from the backseat into the windscreen of a car without backseat constraints, I am always interested in this question.

Some questions:
1. Is the law on child safety seats actually enforced?
2. How much variation is there in the strictness of the law. I note that Australia requires harnesses (attachments from the seat to the car) for child safety seats.
3. Obviously the weight and size of the child are important. You account for this with age, did you look at what happens if the sample is split 2-3 and 4-6?
4. Why would any parent put their child in a car without a car seat, and what type of sample selection bias does this cause?


Great paper Levitt! I will soon have my B.A. in economics, and I was able to read this paper and recall terms from an econometrics class that I have recently taken. Thank you very much for posting it! Aside from that it was interesting to see the data you collected. Based on the topic I would have suspected a different outcome, however the data speaks diferently.

And those are some great questions Chris Adams!


Thanks for publishing this blog. I have been reading it for awhile now and I am hooked.

Various questions regarding automobile safety have interested me for some time, particularly those related to the general question: are SUVs or passenger cars "safer"?

My wife is advocating we buy our 16-yr-old a passenger car because she has seen statistics that imply injuries and fatalities in accidents involving SUVs are higher than in those involving other types of cars. I, on the other hand, have a hard time believing that passengers in an SUV being driven by a responsible individual are in greater danger than those in a Corolla, for example.

There are those who cite statistics from various government sources indicating that injury and fatality rates are higher per mile driven in SUVs than in other passenger cars. This point is made in an article found in a recent Sunday issue of the NY Times.

But this view (based on these statistics) seems to me to have several possible flaws, among them:

(1) the statistics cited talk about auto fatalities involving SUVs, but they aren't clear as to whether the fatalities refer to the passengers of the SUV, or the other vehicles involved in the accident.

(2) it's not clear whether SUVs experience a higher rate of collision than other vehicles (It is possible that the data has "sample bias". For example, drivers of SUVs may be more aggressive or less careful than drivers of other vehicles)

(3) statistics regarding SUV fatalities are largely affected by the incidence of roll-overs, which strike me as operator errors. The circumstances required for a roll-over to occur seem to me to be the result of reckless driving - or - extreme actions taken in an attempt to avert some sort of impending disaster. (But even in those instances in which extreme actions are motivated by an impending danger, there is no guarantee that the same danger that provoked the extreme action causing the roll-over would not have caused some other sort of collision or accident.)

In short, following the "Law of Proportionate Belief" put forth by Arnold Kling, I find it hard for me to believe that a cautious driver abiding speed limits and common sense (and therefore not truly subject to roll-over risk under normal circumstances) but subject to the other dangers of the road (drunk or other dangerous drivers) would be safer in a Corolla than in an SUV.

Perhaps an interesting subject for you to explore some time.



Unless you find significant claims that seatbelts are BETTER than car seats for children 2 and up, the cost of car seats is irrelevant.

Sure, they cost $200. But the cost is a sunk cost when your child turns 2. So, thinking on the marginal level, if there is no difference in safety, and you already own the hunk of junk, why not just leave it strapped in?


Is there anyway to figure out if there correlation to owning the carseat and strapping the child in. in other words if people who have a car seat are more likely to have the child strapped in, possibly caused by just seeing it there empty, or the child is more likely to stay happily strapped in because they can see above the window line and into the front seat. As for the person who above who says that the money is already spent. If there are two cars it would be nice to not have to schlep the safety seat back and forth.


To the blogger with comments with questions about SUVs versus passenger cars, here is a link to an interesting paper by an economist at UCSD that examines many of the issues you raise in your post:

Link to paper on SUV safety

She finds that SUVs are safer for occupants but much more dangerous for the passengers of other cars involved in SUV accidents.

Jim Voigt

So I guess I'm not an economist if I think it's a good idea to spend the $200 on the car seat even if it only on the "chance" that it will do a better job of saving my child's life.

To put it another way, do economists seriously stare at the selection of car seats at Babies-R-Us and wonder to themselves whether the marginal difference in safety is really worth the money? Perhaps this just integrates a little too much reality into the otherwise wholly real-world-detatched science of economics. Ironic considering its claim to present "the world as it really is".

A quick technical question, though. Does your study include kids all the way down to infancy? I assume a child seat is safer than mom holding baby in her lap in the front seat.


DOn't you have to buy a larger seat as the child gets older? Then first seat would be a sunk cost. If the extra seat only had a marginal safety feature, then it may be a waste of money. It would be similar to paying extra for airbags or something.


Jim Voight said: So I guess I'm not an economist if I think it's a good idea to spend the $200 on the car seat even if it only on the "chance" that it will do a better job of saving my child's life.

To put it another way, do economists seriously stare at the selection of car seats at Babies-R-Us and wonder to themselves whether the marginal difference in safety is really worth the money? Perhaps this just integrates a little too much reality into the otherwise wholly real-world-detatched science of economics.


In my mind, the issue primarily relates to mandating the use of car seats via law. Jim, you may not have a problem coming up with $200 to put your child in a car seat. Some other person's economic reality may be very different. $200 may be the difference between paying the electric bill, the child care, or for food--all of which (I assume) would have a demonstrably greater negative impact on the child if they were not paid than would not putting that same child in a car seat (at least according to this study).

Economics is about the allocation of scarce resources. For most people, money is the scarcest of those resources. To the extent that state legislatures mandate that children must ride in a child seat (I believe up to age 8 here in Colorado), it should be a sound policy decision based on logic and a rational decision-making process. Remember, just because it is not legislated that you have to put your child in a car seat, doesn't mean you are prohibited from it. You may look at the economics of the situation and decide that it is a very reasonable allocation of your resources.



On the sidebar question of "Are SUVs or passenger cars safer?" I wonder how much of what makes SUV more hazardous - if even only to the occupants of the other vehicle - if the false sense of security that an SUV-driver is likely to have. That is, given that it seems common for many to equate a big, rugged SUV with increased safety, I wonder if that mindset alone might cause the drivers of SUVs to be bolder or more cavalier in their operating practises...

On the flip side, I'm more inclined to believe that the engineers of smaller passenger cars have to put all that much more effort into designing crumple zones, airbag systems, and other occupant safety elements due to the space contraints and those smaller, seemingly less rugged vehicles actually fare better in collisions. Reviewing the insurance industry's crash-test data on some of the smaller cars on the U.S. roads (VW New Beetle, for example) tends to lend a dash of credibility to this line of reasoning...



Two points:
1. Do any of you have little kids? I think that car seats actually cost $50 or $60.
2. But the real cost of car seats is having to buy a larger car. You can't fit more than two of them into smaller and medium cars. Hence, if you have three kids under age 8 or two kids, one of whom occaisionally has a friend, you are trapped into purchasing a minivan or SUV and the extra gas, insurance etc. to run those.


Chris Adams asked,"Is the law on child safety seats actually enforced?"

I know of one person (my brother-in-law) who was pulled over by a cop because my nephew (7 years old at the time) was not wearing a seatbelt. This was on an island in the state of Maine,USA, all gravel roads, speed limit on the whole island-- 15 mph.


I have no kids but do have relations in the car seat age range and this topic has interested me for a while.

If I'm not mistaken, the literature from the Children's Safety groups makes the claim that in school aged children, it isn't that they won't survive a crash if only in a seat belt, but that they're far less likely to sustain injury in a car seat.

This has always made sense to me: A child in a framework so that the seatbelt fits across his/her lap instead of abdomen would be less likely to sustain internal injuries from the seatbelt.

To me, that fact is enough to make kids sit in car seats.

Hank Weiss, PhD

Glad to see you brought up the non-fatal injury issue and the need to look at a study sample of all crashes in your blog. But it deserves a little amplification.

While individual analyses can be complex, when performing efficacy studies the rationale to look at study samples of all crashes rather than deaths is easy to understand: Successes don't show up in fatality data.

Examining efficacy and cost effectiveness by only looking at fatality data is not ideal for other reasons as well. Preventing non-fatal injuries can be a key part of the equation, especially considering the importance of reducing serious neurological injury (brain and spinal cord).

You will also find some descriptions of useful state-based linked crash data sets if you want to explore these and other related questions from the NHTSA CODES site.

Lastly, it is noted that a somewhat similar issue occurs when measuring vaccine efficacy. As more and more people are immunized, vaccine failures make up more and more of known cases of disease. It does not mean the vaccines are less and less effective. Proportional representation can be misleading without correct and full denominator data.

Bottom line; the bulk of the evidence says to reduce injury, children are safer in car seats.



I don't think that our public resources should be spent on ticketing people who choose not to put their children in car seats, refuse to wear helmets, or insist on doing stupid things unless they are putting others in harms way. I think that we should allow natural selection to take its course because stupid people are taking over the world. There is no reason why I should have to read "caution contents hot" on my coffee cup.


I'm in agreement with the previous poster - let stupid people do their stupid things and get their stupid selves killed. I think of it as a little Clorox in the gene pool. This line of reasoning makes me wonder why we don't just legalize drugs and let the junkies do us all a favor.

However, the big glaring problem with this line of resoning is that stupid people's stupid actions usually have repercussions that extend onto plenty of non-stupid innocent bystanders. Do you want your family or friends living in the house next door when one of those aforementioned junkies, for example, in a drug-induced stupor torches his own house while trying to light his crack pipe?

Back to the topic at hand, child car seats offer a kind of indirect safety as well. They offer the chance for the driver to apply more focus on the task at hand - driving - rather than fussing with kids and thereby causing an accident.



I only looked briefly at the FARS data, but I didn't find their definition of incapacitating. Does it include, say, both people with spinal cord injuries and people with broken legs? I would think medical literature would tend to be more specific.

I agree that it would make economic sense to try to design seat belts that would adjust to properly fit a child. I know it would be far simpler to be able to instantly put my children in any vehicle and have a safe restraint for them. I think it makes sense from a safety perspective as well. When you watch crash tests with harnessed child-safety seats (3 yr old in car seat), you see the child-safety seat moves forward from the car seat and the child moves forward from the child-safety seat. It seems forward movement could be reduced by restraining with seat belts alone. Of course, seat belts would need to be designed to actually fit correctly and comfortably on a child.

I appreciated that the Times article encouraged parents to always use seat belts or child-safety seats. However, the biggest difference in fatalities I saw among restrained children was between properly-used restraints and improperly-used restraints. It would have been more helpful and accurate to emphasize that parents should always make sure their kids are correctly restrained whatever system they are using.

It is true that both seat belts and child-safety seats can be used incorrectly. The big difference in my mind is who controls correct usage. With a harnessed child-safety seat, the parent is responsible to correctly install it and harness the child. While toddler escape-artists do exist, most harnessed seats do not require much from the child. (They can even fall asleep and still be properly restrained in most instances.) With current seat belts, used alone or with boosters, the child is the one who is mainly responsible for using it correctly. A parent may buckle them in and try to constantly monitor their positioning, but that is hard to do when driving. (Probably not too safe, either.) So with what's currently available, I would say the real question is 'Who are you going to trust your two-year-old's safety to?' (And I personally wouldn't trust the DVD player.)


Stephen J. Dubner

A reader named Thomas J. Bzik sent us the following comment:

My first professional job as a statistician was with the Center for the Environment and Man. They were contracted to analyze accident data for NHTSA to retrospectively estimate the benefits obtained from the FMVSS rules (mostly the 1973 rules). he
analyses were being performed in the 1978-1980 time window (mostly 1979). One of the items for
evaluation was child safety seats. We evaluated child seat performance based on the child seats implicit to usage in that era and found that car seat belts were more effective for children and infants. NHTSA was incensed by this conclusion and we had to reanalyze the data numerous times and ways eliminating parts of the data they decided they did not like for, in our opinion, questionable reasons. No matter how the data was sliced and diced we found that car seat belts were more effective for children
and infants (highly statistically significant effects). NHTSA did not like this conclusion and
obviously never went public with it (did not want to discourage child seat usage despite the relative ineffectiveness ompared to existing seat belt systems). In fact they verbally indicated that they were not even going to take the results to the child
seat manufacturers of the day to foster future improvement. NHTSA felt that the manufacturers
would be "upset" by the conclusions. We felt that
the child seat manufacturers richly deserved to be upset! Ever so slightly to NHTSA's credit is that subsequently got around to revising the child safely seat standards in 1982.


Jim Voigt

The flaw in the theory of scarce resources as it relates to child seats is that there are several charities that provide child seats for parents who cannot afford them. Also, at least in my neighborhood, $200 represents about the top end of car seats. We didn't pay even close to that for ours. We are also not so lazy as some other commenters to approach the daunting task of moving a car seat from one car to another as so overwhelming as you bypass this even marginal increase in safety for our child. Was that person serious?

Apparently this was not part of your otherwise somewhat slightly almost thorough research.

I suppose the main issue I have with your style of statistical analysis is that you so commonly focus on data that appears to prove a point and run with it without truly examining other simple factors that may very realistically affect your conclusions. A=B, B=C, so A=C works on a laptop, but doesn't always describe the real world.

Your concern for the validity of your sample selection on this study appears to be unique, as I've identified several "conclusions" of yours that appear to have used either flawed, incomplete or even ridiculous sample selection methodology.

Again, I'm no economist. But perhaps it takes a non-economist to point out some of the very simple common sense ways your various analytical techniques fall short of presenting the world as it really is regardless of your claims of doing exactly that.

Makes for an entertaining website and book, but that's about where its usefulness ends.



I want to underline a (much) earlier point; the fact that Pennsylvania enforces car/booster seat use until age *8* has the effect of forcing families into larger cars. A Subaru Forester ought to be a sensible family vehicle. In the old days, such a vehicle would hold three people (kids, or two adults and a kid) in the back seat. Both car and booster seats are wider than many adults, let alone kids. So if you have three children under the age of 8 (and I'm not even sure they're allowed in the front seat yet at that age), you are pretty much required by law to drive a minivan or SUV.
The price of non-compliance is a fairly steep ticket, and I *think* if you have an accident and the child is not in the correct seat, you can be charged with child endangerment.